Down Ratio

The share of tickers in the study universe whose latest AI prediction points down; the core bearish breadth statistic in SD clustering.

The Down Ratio is a bearish market breadth statistic used in technical analysis and machine learning models to measure the proportion of assets in a given universe that are predicted to decline in price. It is calculated by dividing the number of individual tickers with a downward price prediction by the total number of tickers that received any prediction. This metric helps investors gauge the overall bearish sentiment across a broad basket of stocks, rather than relying on a single index or asset. In systematic clustering models, the Down Ratio serves as a primary filter for identifying market-wide downward trends. For a specific trading day to be classified as bearish or to trigger a downward signal, the Down Ratio must exceed a statistically significant threshold. Because predictive models often exhibit structural biases or lean cautious, the historical average for this ratio may sit well above 50%, meaning the thresholds required to signal a true bearish cluster are typically set much higher than a simple majority. Investors monitor the Down Ratio to identify periods of widespread market vulnerability. A high Down Ratio indicates broad-based weakness, suggesting that a market decline is being driven by many participating stocks rather than a few heavily weighted laggards. Conversely, a low Down Ratio suggests that very few stocks are expected to fall, which can signal underlying market strength or a potential bullish turn. One common pitfall is interpreting a Down Ratio slightly above 50% as an active sell signal; without comparing the ratio to its historical baseline or standard deviation thresholds, investors may misinterpret normal market noise as a systemic downturn.

In a curated study of 500 large-cap stocks, a machine learning model generates price predictions for all of them. The model predicts that 390 of these stocks will decline in price over the next target period, while the remaining 110 are predicted to rise. The Down Ratio is calculated as 390 / 500 = 0.78, or 78%. If the historical 2 standard deviation threshold for this study is 78.6%, this result falls just short of triggering a high-conviction bearish clustering signal.